A price based spectrum sharing scheme in wireless cellular networks

Journal of Information Engineering and Applications www.iiste.org
ISSN 2224-5782 (print) ISSN 2225-0506 (online)
Vol.3, No.8, 2013
1
A Price Based Spectrum Sharing Scheme in Wireless Cellular
Networks
Sadath Kalolo (Corresponding author) 1
*, Anael Sam1
, Dina Machuve1
and Justinian Anatory2
1
The Nelson Mandela African Institute of Science and Technology, P.O Box 447 Arusha, Tanzania
Emails: *kalolos@nm-aist.ac.tz, anael.sam@nm-aist.ac.tz, dina.machuve@nm-aist.ac.tz
2
University of Dodoma, P.O Box 259, Dodoma, Tanzania
Email: anatory@engineer.com
Abstract
Radio frequency spectrum scarcity has become a high priority research area over the past few years. The huge
increase of network subscribers with multimedia applications coupled with underutilization of radio frequency
spectrum motivates the search for other measures to address the scarcity of radio frequency spectrum. This work
investigates on a price based spectrum sharing scheme for connection-oriented traffic in wireless cellular
networks as a solution to address the scarcity of radio frequency spectrum. Dynamic pricing approach is applied
with traffic overflows into neighbor networks. Performance evaluations of the scheme at steady state using
MATLAB simulations reveal significant gains to the quality of service. Application of the scheme to highly
loaded network traffic improves both network revenue and traffic channel utilizations.
Keywords：：：：Pricing, spectrum sharing, traffic overflows, Quality of service, channel utilizations, Wireless
cellular networks.
1. Introduction
The wireless cellular industry is growing fast coupled with an increase in bit rates and multimedia applications.
The global mobile cellular subscriptions stand at 6.8 billion users in 2013 with a global penetration rate of 96%,
128% in developed countries, and 89% in the developing countries(ITU 2013). In Tanzania, the mobile cellular
subscription was about 27.4 million in December 2012 with a penetration rate of 61%(TCRA 2012). This growth
demands for extra radio frequency spectrum capacities especially during busy hour periods.
Various approaches are used to optimize radio frequency utilization in wireless cellular networks including cell
split and frequency reuse, overlapping of cell layers, and dynamic channel allocation techniques (Rappaport
2009), (Garg 2007) &(Katzela and Naghshineh 1996). However, even with the high level of frequency reuse, as
with the introduction of micro, pico and femto cells, during busy hour periods the networks are overwhelmed
with high volumes of traffic resulting into congestions. Furthermore, repeated use of most of these methods ends
up with network complexity and an increase in radio frequency interference.
Wireless cellular network traffic is heterogeneous with spatial and temporal fluctuations. The demand for
wireless radio resources differs significantly between peak and off-peak hours. Radio frequency spectrum
utilization varies in the range of 15% to 85% in the band below 3 GHz (Akyildiz, Lee et al. 2006). Currently,
most wireless cellular networks offer cheaper or even free off-peak calls as a marketing incentive, in an attempt
to effectively utilize the available network capacity. However, this approach lacks flexibility to take into account
the actual network load by increasing the price when higher demands are anticipated (Bouroche 2003). Variation
of prices according to utilization assists to lighten network congestion and potentially yields more revenue to
network operator (Fitkov-Norris and Khanifar 2000)&(Viterbo and Chiasserini 2001).
Radio frequency (RF) spectrum sharing exploits the differences in traffic loadings between networks (Heinonen,
Pirinen et al. 2008). Sharing are feasible in three different cases depending on the objective; Sharing as a last
option, always connected to the base station that has the greatest of unused resources, and as a secondary user.
For networks with primary and secondary traffic routes, extra capacities are obtained by allowing incoming
traffic to overflow from primary channel groups to secondary channel groups (Iversen 2005). Alternative path
routing of overflow traffic widely improves the performance of hierarchical networks and IP networks (Huang,
Ko et al. 2008).
In Tanzania, wireless cellular networks’ traffic is disrupted by frequent power cut-off due to power shortage and
rationing. The growth of the industry, use of multimedia applications and frequent power rationing motivates the
search for other RF management strategies to address the problem of RF scarcity and quality of service. This
work investigates a price based spectrum sharing scheme for connection- oriented traffic in wireless cellular
network that applies dynamic pricing strategies with traffic overflows into neighbor networks at busy hour
periods. The rest of the paper is organized as follows; section II presents the state of art of RF spectrum
optimization and utilizations. Section III presents the system and traffic model of the network. Section IV
presents the network simulations and the results. Section V concludes our work paving directions for further
research.

Vol.3, No.8, 2013
2
2. State of Art
RF spectrums are assigned to wireless cellular networks exclusively on long term basis. In Tanzania seven
companies are licensed to offer wireless cellular network services(TCRA 2011), four of which operates GSM
technology while the remaining operate CDMA technology services. Absolute RF Channel Numbers (ARFCNs)
are assigned to GSM operators both in the 900 MHz and 1800 MHz bands. Furthermore, all GSM operators are
also licensed with UMTS/WCDMA services in 2100 MHz band. Wireless cellular network operators implement
a combination of traffic channels from the above frequency bands to support multi band terminals. Mobile
networks evolve from macro cell systems to micro, pico and femto cell systems in the search for more capacity
to accommodate more network users. Two categories of channel allocations are noted to be in use currently; the
baseband hoping (fixed channel assignment) and frequency hopping (dynamic channel assignment). Baseband
hoping are mainly used in rural areas sites are far from each other and each radio transceiver is assigned a fixed
frequency while frequency hopping schemes are commonly used in towns and cities. Cells in rural areas are
grouped into bigger clusters with a 4 × 3 cell cluster being in common use. In cities and towns a tight frequency
reuse pattern is used due to high traffic loads in these areas with a 1 × 1 cell cluster being in common use. To
avoid frequency collision, RF hopping techniques are used with appropriate parameter settings including; the
mobile allocation (MA), hopping sequence numbers (HSN), and mobile allocation index offset (MAIO).
However, some of the cells are poorly utilized while others are over utilized. Table 1 presents some of traffic
channel utilizations per cell extracted randomly from a wireless cellular operator as observed in our study.
Table 1: Traffic channel utilizations per cell
S/N
Sample Traffic Channel (TCH) Utilization per cell
Site No. Average TCH utilization (%) Site No. Average TCH utilization (%)
1 Site 1 122.4 17 Site 17 58.6
2 Site 2 6.6 18 Site 18 206.2
3 Site 3 19.3 19 Site 19 197.3
4 Site 4 115.2 20 Site 20 152.2
5 Site 5 118.7 21 Site 21 64.1
6 Site 6 91.3 22 Site 22 4.3
7 Site 7 131.4 23 Site 23 78.0
8 Site 8 106.4 24 Site 24 145.5
9 Site 9 114.1 25 Site 25 60.0
10 Site 10 154.4 26 Site 26 43.1
11 Site 11 138.6 27 Site 27 45.5
12 Site 12 73.6 28 Site 28 67.8
13 Site 13 142.2 29 Site 29 3.6
14 Site 14 103.7 30 Site 30 184.0
15 Site 15 74.8 31 Site 31 129.2
16 Site 16 160.4 32 Site 32 237.1
3. System Model
We consider two wireless cellular networks each operating the same radio access technology to allow RF
spectrum sharing when such requirement arises. Each network operator is basically assigned with a certain
amount of absolute RF channel numbers (ARFCNs). The home network and the neighbor network together
constitute a n + n server/ traffic channels loss system (M/M/n/n queue), where n = n + n . Mobile stations
are randomly distributed in the area of study and are randomly assigned to their home network from which they
buy services regularly. Sequential (ordered) hunting is done when mobile stations arrive at the base transceiver
stations such that home network traffic channels are always utilized first. When incoming traffic load exceeds the
capacity of the base transceiver station, extra traffic load overflows into neighbor wireless cellular network. This
depends on the availability of traffic channels at that operator.
We assume a pure chance traffic type I is offered into a system that works as a loss system (Blocked calls are
cleared). The call arrivals follow the standard poison modulated process with mean arrival rates of λ and
service times that are exponentially distributed with mean service rate of µ . Fig. 1 presents a two dimensional
Markov Chain used to represent the system evolution. Fig. 2 presents a traffic model where blocked calls in the
home network overflows into a neighbor network with available traffic channels n . Calls that overflow into the
neighbor network do not return to their home network even when the traffic channels are released. Overflowing
traffic is not pure chance traffic type I any more but burst. If all available traffic channels in the neighbor
network are also occupied, arriving calls are blocked and cleared.

Vol.3, No.8, 2013
3
Figure 1: A two dimensional Markov Chain
Figure 2: Traffic model
1.1 A price based spectrum sharing algorithm
We propose a scheduler that is decentralized and works at the base station controller/radio network controller.
The scheduler iterates through the base transceiver stations assessing their traffic channel utilization status, thus
providing base information for price adjustments based on the current traffic load. When traffic load is low, the
price is lowered to motivate more network users to place their calls into the network. As traffic load increases
arrangements are made such that when the home network capacity is used up, the extra traffic load overflows
into traffic channels available at the neighbor network. The limitation of the traffic channels at the neighbor
network operator will cause a steady increase in call pricing accordingly to deter additional traffic and suite the
available capacity.
1.2 The demand function
We use the demand function appearing in (Viterbo and Chiasserini 2001), where in (1) p is the call price per
time unit. Q is the quality of service index for call success probability (Q = 1 − P ) and P is the call blocking
probability. The parameters α and β are related to the user population characteristics.
, ) = ! "#$%&)
(1)
The demand in the time span ∆t is assumed to equal the actual traffic generated in the time span of ∆t, thus,
∆). ! "#$% *+)
= ∆). ,. 1 − -.) (2)
Where, , =
/0
10
, for 0 ≤ i ≤ n − 1
89 =
:0
*+
. ! "#0$% *+)
(3)
for 0 ≤ i ≤ n − 1

Vol.3, No.8, 2013
4
Assuming that p = p;, p , p , … . p= ) is the price vector that represents the cost per time unit of a call started
when the system is in a given state. Network users react to the changes in price per time unit by reducing their
number and duration of calls. Significant reaction is expected when network users access the RF traffic channels
from the neighbor network since these traffic channels are charged at slightly higher prices than the charges of
home network. If t> = and t>?@ are the minimum and maximum call duration of a call per time unit, then the
call duration in state i, 0) is given by (4).
:0
= ):AB! CD #0 #E)
(4)
for 0 ≤ i ≤ n − 1)
Where,
K> = ln
t>?@ t> =⁄ )
p= − p;)
The average call termination rate when the system is in state i, 0) is obtained by (5).
µ9 = I; + I + ⋯ + I9 (5)
for 1 ≤ i ≤ n
1.3 Call blocking
Analysis of the two dimensional Markov Chain is limited by state explosions as the number of traffic channels is
increased. We therefore obtain the blocking probability for the home network is at steady state and approximate
the overflow traffic using Fredericks and Hayward’s approximation method. Considering a single dimension
markov chain with n + 1 states and using cut equations we obtain the call blocking for the home network as in
(6).
p n ) = E n , a) = P = L∑ ∏
OPQR
1P
S
T
=R
ST; U ∏
OPQR
1P
=R
T (6)
The overflow traffic is given by (7).
w = a. E n , a) =
OPQR
1P
L∑ ∏
OPQR
1P
S
T
=R
ST; U ∏
OPQR
1P
=R
T 7)
The mean and peakedness of overflow traffic are obtained by (8) below (Iversen 2005).
Mean, w =
/0QR
10
L∑ ∏
/0QR
10
X
9T
YR
XT; U ∏
/0QR
10
YR
9T
Peakedness, z = 1 − [
OPQR
1P
L∑ ∏
OPQR
1P
S
T
=R
ST; U ∏
OPQR
1P
=R
T +
]PQR
^P
=R$ [
]PQR
^P
$_
]PQR
^P
`∑ ∏
]PQR
^P
a
PbR
cR
abE
d
QR
∏
]PQR
^P
cR
PbR
e
(8)
We apply Fredericks and Hayward’s approximation method (Iversen 2005) to approximate the overflow traffic
blocking as in (9).
E n, a, z)~E
=
g
,
?
g
, 1)~E
=
h
,
?
g
) (9)
Thus, the overflow traffic blocking is given by (10).
wi =
OPQR
1P
L∑ ∏
OPQR
1P
S
T
=R
ST; U ∏
OPQR
1P
=R
T E j
=k
hk
,
l
hk
m (10)
And system traffic congestion (STC) is obtained by (11).
STC = `∑ ∏
λPQR
µP
S
T
=R
ST; d ∏
λPQR
µP
=R
T E j
=k
hk
,
l
hk
m (11)
1.4 Traffic channel utilizations
We assume a sequential assignment of traffic channels to incoming traffic. The traffic carried by the iqr
channel
is the difference between the traffic lost from i − 1)qr
channel and the traffic lost from iqr
channel. Traffic
carried by the iqr
channel is independent of the number of channels after the iqr
channel in the hunting order. In
the home network traffic channel utilizations is obtained by (12).
s9 =
/0QR
10
L∑ ∏
/0QR
10
X
9T
YR
XT; U L∏
/0QR
10
X
9T − ∏
/0QR
10
X
9T U (12)
for: 0 ≤ k ≤ n
For the spectrum shared network the traffic channel utilizations when the system is in state i, j) is given by (13).
s9 =
/0QR
10
L∑ ∏
/0QR
10
X
9T
YR
XT; U w j
Yk
xk
,
y
xk
m L∏
/0QR
10
X
9T − ∏
/0QR
10
X
9T U (13)
For 0 ≤ z ≤ { and 0 ≤ | ≤ {
1.5 Network revenue
We assume that g = g , g , … , g~, … . g=R
) is the revenue vector representing the total revenue per time unit of
the network when a system is in a given state. The revenue per time unit when the system is state i, 0) is given

Vol.3, No.8, 2013
5
by (14).
•9 = ∑ 9
9
; for 0 ≤ i ≤ n (14)
And the total network revenue is obtained by the following equation.
€9 = ∑ 1 − -•9)9
9T; ∗ •9 (15)
for 0 ≤ i ≤ n
In a spectrum shared network when the system is in state i, j), the total network revenue is obtained using (16).
€9 = ∑ 1 − ƒ„…) ∗ •9
9
9T; (16)
for 0 ≤ i ≤ n
4. Simulations
We apply linear pricing vector (dynamic prices) to the demand function and allow extra incoming traffic load to
overflow into a neighbor network. We further assume that eight traffic channels (TCHs) are available from the
home network cell and traffic channels from the neighbor network are available in intervals of one traffic
channels. We also assume the variation of call holding periods from a maximum of 180 seconds to a minimum of
60 seconds. The ratio of the constants β and α in the demand function is assumed to equal four (
β
α† = 4) as used
in (Viterbo and Chiasserini 2001). We allow the dynamic price to vary from 0.7 to 1.4 units per second and for
the purpose of simplification, in all experiments we ignore the hand-off traffic. We use MATLAB simulations to
calculate the probability of blocking iteratively until the steady state is reached. The parameters obtained at
steady state are then used to compute various performance measures of the scheme.
4.1 Simulation Results
Dynamic pricing with spectrum sharing schemes produces lower call blocking probability of incoming calls
when traffic overflows into a neighbor network. Fig. 3 presents the comparative results of call blocking
probabilities (system traffic congestions) for a dynamic pricing with and without spectrum sharing.
Figure 3: Call blocking for a dynamic priced network with and without spectrum sharing
Traffic overflow causes poor traffic channel utilization in the system. This is explained by the burstness of the
traffic that overflows into the neighbor network (Peakedness = Variance Mean⁄ being greater than one).
However, when traffic load is high the traffic channel utilizations for dynamic prices with spectrum sharing
exceed dynamic prices without spectrum sharing. This suggests the applications of the scheme to highly traffic
loaded network scenarios. Fig. 4 presents the simulations results for traffic channel utilization.
10
-0.4
10
-0.2
10
0
10
0.2
10
-12
10
-10
10
-8
10
-6
10
-4
10
-2
10
0
p (Traffic Load)
Pb(BlockingProbability)
Call blocking - dynamic pricing
without sharing
Sharing, N2=1
Sharing, N2=2
Sharing, N2=3

Vol.3, No.8, 2013
6
Figure 4: Traffic channel utilization for a dynamic priced network with and without spectrum sharing
A small increase to the total network revenue is observed when dynamic pricing with traffic overflows into the
neighbor network is applied as compared to dynamic pricing without spectrum sharing. Fig. 5 presents
simulation results for the total network revenue in the network. However, the change in the network revenue
generated for successive increase of neighbor network traffic channels is minimal.
Figure 5: Total network revenue for a dynamic priced network with and without spectrum sharing
5. Conclusions
We have presented a price based spectrum sharing scheme for connection oriented traffic in wireless cellular
networks. Dynamic prices are applied with traffic overflows into neighbor networks. Performance evaluation
using MATLAB simulations reveals significant gains in the quality of services offered and an improvement of
traffic utilizations when sharing are opted in a highly traffic loaded network. Gain in the network revenue is also
observed when spectrum sharing is allowed. The application of the scheme to the connectionless oriented traffic
is left as a direction for further research.
References
Akyildiz, I. F., W. Y. Lee, et al. (2006). "NeXt generation/dynamic spectrum access/cognitive radio wireless
networks: a survey." Computer Networks 50(13): 2127-2159.
Bouroche, M. (2003). Meeting QoS Requirements in a Dynamically Priced Commercial Cellular Network,
Department of Computer Science, University of Dublin, Trinity College.
Fitkov-Norris, E. and A. Khanifar (2000). Evaluation of dynamic pricing in mobile communication systems,
Citeseer.
Garg, V. K. (2007). Wireless communications and networking, Morgan Kaufmann.
Heinonen, V., P. Pirinen, et al. (2008). Capacity gains through inter-operator resource sharing in a cellular
network.
Huang, Q., K.-T. Ko, et al. (2008). "Approximation of loss calculation for hierarchical networks with
multiservice overflows." Communications, IEEE Transactions on 56(3): 466-473.
ITU. (2013). "ICT Facts and Figures." Retrieved 21st
May 2013, from http://www.itu.int/en/ITU-
D/Statistics/Documents/facts/ICTFactsFigures2013.pdf.
10
-0.4
10
-0.2
10
0
10
0.2
10
-10
10
-8
10
-6
10
-4
10
-2
10
0
p (Traffic Load)
TrafficChannelUtilization
Traffic Channel Utilization - dynamic pricing
Without sharing
Sharing, N2=1
Sharing, N2=2
Sharing, N2=3
10
-0.4
10
-0.2
10
0
10
0.2
10
0
10
1
p (Traffic Load)
R(NetworkRevenue)
Total network revenue - dynamic pricing
without sharing
with spectrum sharing N2 =3

Vol.3, No.8, 2013
7
Iversen, V. B. (2005). "Teletraffic engineering and network planning." COM, DTU.
Katzela, I. and M. Naghshineh (1996). "Channel assignment schemes for cellular mobile telecommunication
systems: A comprehensive survey." Personal Communications, IEEE 3(3): 10-31.
Rappaport, T. S. (2009). Wireless Communications: Principles and Practice. Upper Saddle River, NJ 07458,
Prentice Hall PTR.
TCRA. (2011). "Licensed Cellular Network Operators." Retrieved 28th
May 2013, from
http://www.tcra.go.tz/index.php/licensing/licensed-operators/2-tcra/39-network-services-licences.
TCRA. (2012). "Telecommunications Statistics December 2012." Retrieved 21st
May 2013, from
http://www.tcra.go.tz/images/documents/telecommunication/TelCom%20Statistics%20Dec%202012(Fi
nal).pdf.
Viterbo, E. and C. F. Chiasserini (2001). Dynamic pricing for connection-oriented services in wireless networks,
IEEE.

This academic article was published by The International Institute for Science,
Technology and Education (IISTE). The IISTE is a pioneer in the Open Access
Publishing service based in the U.S. and Europe. The aim of the institute is
Accelerating Global Knowledge Sharing.
More information about the publisher can be found in the IISTE’s homepage:
http://www.iiste.org
CALL FOR PAPERS
The IISTE is currently hosting more than 30 peer-reviewed academic journals and
collaborating with academic institutions around the world. There’s no deadline for
submission. Prospective authors of IISTE journals can find the submission
instruction on the following page: http://www.iiste.org/Journals/
The IISTE editorial team promises to the review and publish all the qualified
submissions in a fast manner. All the journals articles are available online to the
readers all over the world without financial, legal, or technical barriers other than
those inseparable from gaining access to the internet itself. Printed version of the
journals is also available upon request of readers and authors.
IISTE Knowledge Sharing Partners
EBSCO, Index Copernicus, Ulrich's Periodicals Directory, JournalTOCS, PKP Open
Archives Harvester, Bielefeld Academic Search Engine, Elektronische
Zeitschriftenbibliothek EZB, Open J-Gate, OCLC WorldCat, Universe Digtial
Library , NewJour, Google Scholar

A price based spectrum sharing scheme in wireless cellular networks

More Related Content

What's hot (19)

Similar to A price based spectrum sharing scheme in wireless cellular networks (20)

More from Alexander Decker (20)

Recently uploaded (20)

A price based spectrum sharing scheme in wireless cellular networks